 

 

THE INFLUENCE or GRAMMATICAL TRANSI-OkM
IN A SYLLOGISTIC REASONtNG TASK

Thesis for the Degree of M. A.
MICHIGAN STATE UNEVERSITY
Marcia Z. Lippman
196:6

1142515

 

 
 
  

w w .

1.18» {R r '

“digit Stab
Uniw try

 
 
   

ABSTRAC T

THE INFLUENCE OF GRAMMATICAL TRANSFORM
IN A SYLLOGISTIC REASONING TASK

by Marcia Z. Lippman

Recent research in verbal behavior has been based on the
assumption that there may be a correSpondence between syntactic
relations as described by modern linguists and the psychological
Operations performed by the language user. The differential in-
ﬂuence of syntactic structures (active, passive, negative, question,
etc.) has been reﬂected in behavioral measures such as reaction
time, motor re3ponse generalization, and recall, with sentenial
material. The active-affirmative structural form has been re-
garded as psychologically primary as well as syntactically primary.
Some evidence suggests that §s reword syntactically complex sen-
tences into the active-affirmative at some stage in processing.

The present experiment investigated the inﬂuence of
syntax on the operations of the language user by varying the
structure of the major premise of syllogistic reasoning problems.
Four behavioral measures were used: solution latency, rated

difficulty, number of correct solutions and structure of verbal

Marcia Z . Lippman

solution. Of particular interest was the Passive -Active comparison;
of secondary interest the Negative -Affirmative comparison.

It was hypothecized that (a) problems written in the Passive
should take longer to solve and be rated more difficult than those
written in the Active; (b) problems written in the Negative should take
longer to solve and be rated more difficult than those written in the
Affirmative; (c) verbal solutions should be given predominantly in
the Active form; (d) verbal solutions to Passive problems may be
given in the Passive to difficult problems as an atmosphere response.

A set of 32 syllogisms, each having a different semantic
content, was employed. A valid conclusion could be reached for half
of the syllogisms (determinant items); no valid conclusion could be
reached for the other half of the syllogisms (indeterminant items) .
Eight of the determinant syllogisms contained a Negative term in the
major premise; eight of the indeterminant syllogisms contained a
Negative term in the minor premise. Voice (Active, Passive) and
Position ( l, 2) of the major premise of each syllogism was varied.
Four forms were prepared such that each of the 32 syllogisms appeared
on each of the four forms but represented a different combination of
variables (Voice and Position) on each. In this way semantic content
was counterbalanced. The two premises for each problem were manu-
ally presented on 4 x 5 white index cards. § was handed each of the

32 cards in succession by E. §_ was instructed to read the premises,

Marcia Z . Lippman

reach a solution, and tell it to E. Solution latency and the verbal

solution were recorded. § rated each problem on an 11-point dif-

ficulty scale printed on a card in front of him after giving the solution.
§_s were 64introductory psychology students.

The hypotheses were confirmed. More solution time was
required when problems were written in the Passive or Negative than
in the Active or Affirmative. The Passive and Negative problems
were also rated as more difficult. Verbal solutions were given pre-
dominantly in the Active (85% to determinant items and 61% to
indeterminant items) rather than Passive. (evidence for rewording
of Passive) With respect to predictionDadichotomy of _S_s was found.
One grouping of _S_s took less time to give a Passive than Active re-
Sponse to Passive problems and rated those problems to which they
gave a Passive response as easier. The second grouping of §S took
more time to give Passive than Active responses to Passive problems
and rated those problems to which they gave a Passive reSponse as
harder. This latter group gave significantly more Passive responses
than did the former. It was suggested that perhaps the dichotomy of
_S_s represents those _S_s who gaye Passive responses to easy problems
because the solution was so obvious for them that no rewording into
the Active was necessary, and those SS who gave the Passive reSponse.
to )difficult problems as an atmOSphere reSponse out of uncertainty or

cautiousness. Two steps may have been involved in the latter _S_s‘

Marcia Z. Lippman

formulation of a solution accounting in part for the additional solution

time required--rewording into Active, then back into Passive.

3 ’ , ' 7 '
Approved: (ZZZ/Z 1. up 27¢ 2551,463/k

Committee Chairman

Date: 8/3/3147 5’

THE INFLUENCE OF GRAMMATICAL TRANSFORM

IN A SYLLOGISTIC REASONING TASK

BY

Ma-r c ia Z. Lippman

A THESIS

Submitted to
Michigan State Univeristy
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1966

ACKNOWLEDGMENTS

The author wishes to thank her thesis committee, Dr.

A. M. Barch, chairman, Dr. M. R. Denny, and Dr. G. I. Hatton.

ii

TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS.................. ii

LISTOFTABLES...................... iv
LISTOFFIGURES.................... v

LIST OF APPENDICES 0 O O O O O O O O O O O O O O O C 0 Vi

INTRODUCTION..................... 1
METHOD 18
RESULTS........................ 22
DISCUSSION....................... 36
SUMMARY 49
REFERENCES. .. .. .. .. .. .. .. .. .. .. . 52

APPENDICES...................... 54

Table

LIST OF TAB LES

Mean solution latency and difficulty ratings for
Active-Affirmative, Active-Negative, Passive-
Affirmative, and Passive-Negative items-

Summary of the analysis of variance of deter-
minant-item solution latencies .

Summary of the analysis of variance Of indeter-
minant-item solution latencies

Summary of the analysis of variance of deter-
minant-item difficulty ratings

Summary of the analysis of variance of indeter-
minant-item difficulty ratings .

Summary Of the results of latency and rating
analyses of variance .

Mean number Of correct solutions .
Percentage Of reSponses given in Active, Pas-

sive, Affirmative, Negative, NO Conclusion
for each combination of variables

iv

Page

29

29

30

30

31

31

32

33

LIST OF FIGURES

Figure
l. Syntactic relations of the eight sentences
2. Mean determinant-item latency as a function Of
Voice, Negation, and Position .
3. Mean determinant-item difficulty rating as a

function Of Voice, Negation, and Position

Page

34

35

Appendix

LIST OF APPENDICES

Instructions
Difficulty Rating Scale

Syllogisms. . . .

vi

Page

54
56

57

INTRODUCTION

With the recent interest in the acquisition and recall of
sentential material, given impetus by Miller's 1962 EPA Address,
the role Of syntactic as well as semantic components in these Oper-
ations has been recognized and explored in some detail. The role
of meaning in the learning of verbal material has been repeatedly
demonstrated in the psychological literature. The semantic com-
ponent (meaning), however, is only one asPect of a sentence. The
first investigations of the syntax component Of a sentence were pri-
marily concerned with demonstrating that structure or syntax was
important in learning and perception.

A modification Of the procedure used by Miller and
Selfridge ( 1950) of varying the order Of approximation to English
was adopted for these demonstrations. Miller and Selfridge had
shown that recall Of strings Of words improved as their order of
approximation to the statistical pattern of English increased. Their
procedure, however, did not distinguish between syntactic and
semantic factors. Verbal context was defined as the extent to which
the choice Of a particular word depended upon the word preceding

it, or statistically, as dependent probabilities.

Salzinger, Portnoy, and Feldman (1962) demonstrated
the contribution of both semantic and syntactic. factors to the depend-
ence Of a word upon the preceding one. _S_s were required to guess
words that had been deleted from passages varying in order of ap-
proximation to the statistical structure of English sentences. The
proportion Of words given in the correct grammatical category in-
creased with order Of approximation. This demonstration Of re-
Sponse Specification with increasing orders Of approximation could
be explained in part by the inﬂuence of syntactical structure and in
part by the contextual constraints on meaning.

Materials in which semantic and syntactic rules could
be independently violated were constructed by Miller and Isard
(1963) and by Marks and Miller (1964). The materials were based
on a linguistic rather than statistical analysis of English sentences.
The procedure made it possible to separate the role Of semantic
and syntactic components in the speech perception and recall tasks
employed. Miller and Isard hypothesized that Speech perception
should be most difficult when linguistic rules have been violated
and easiest when Speech follows linguistic rules, allowing for pro-
cessing in habitual fashion. The hypothesis was confirmed. Gram-
matical sentences (semantic and syntactic components maintained)
were most accurately repeated by S5; ungrammatical sentences

(semantic and syntactic components destroyed) were least accurately

repeated; while, anomalous sentences (syntax only maintained) fell
in between. Perceptual processing, thus, was facilitated when sen-
tences followed grammatical rules, whether semantic or syntactic.

Marks and Miller also reported a facilitative effect; of
free recall when sentences followed grammatical rules. Normal
sentences (semantic and syntactic components maintained) were
most accurately recalled; strings of words following no grammati-
cal rules were least accurately recalled; anagram (semantic com-
ponents maintained) and anomalous (syntactic components maintained)
strings fell in between. Both semantic and syntactic components
had facilitative effects on recall.

That §S do in fact, recognize the degree of grammatical-
ness of material has been shown by Coleman ( 1964) . _S_s ranked
materials, which had been generated so that there were four levels
of grammaticalness, in the predicted order. Learning was found to
be a function of the grammatical level.

Once the influence Of syntax on verbal learning has been
demonstrated, psychologists' research interest turned toward in-
vestigating the "psychological reality” Of the generative grammar
rules prOposed by modern linguists: phrase-structure rules and
transformation rules. That is, psychologists have used these rules
as a starting point for investigating the inﬂuence of structure on the

Operations performed by the user of a language. The assumption

was that behavioral measures would reflect a correspondence be-
tween psychological operations and phrase-structure and/or trans-

formational relationships furnished by the grammars (Miller, 1962).

Phrase -Structure Rules

 

Phrase-structure rules apply to the constituent units
(phrases) Of a sentence such that a composite symbol representa-
tive Of a functional unit, e.g. , “the boy" (noun phrase, NP) or
"caught the fish" (verb phrase, VP) is rewritten in a systematic
fashion as follows:

Sentence (S) is rewritten as (--> ) NP + VP

NP ——> T + N

VP —-> v + NP

NP —) T + N

T —') the, a, . . .
N 9 boy, fish, .
V % caught, . . .

The rules governing the expansion Of a composite sym-
bol into a set of symbols are to be applied whenever the composite
symbol occurs. Thus, simple sentences such as "The boy caught
the fish" may be described by formal statements and may be generated
by the application of the phrase-structure rules. Generation would

proceed in reverse order to the descriptive rewriting outlined above.

Two psychological studies have demonstrated that the
constituent units (phrases) Of sentences may inﬂuence _S_s' percep-
tion and learning of sentential material. Johnson ( 1965) has sug-
gested that pharses are not only grammatically functional units,
but, correspondingly, psychologically functional units, and may
be thought Of as integrated response units. The conditional prob-
ability Of an S's recalling a whole phrase correctly, given that one
word of the phrase was correctly recalled, was found to be greater
than if each word were treated as an independent event. A greater
probability Of error at phrase boundaries than within phrases oc-
curred as predicted. Fodor and Beven ( 1965) provided additional
support for considering the phrase as an integrated response unit.
§S reported the location Of clicks heard during sentences as having
occurred at the phrase boundaries. Such di8placement Of clicks
toward phrase boundaries was interpreted as Ss' way of insuring

the integrity of these units.

Transformation Rules

 

A second set of generative grammar rules are the trans-
formation rules proposed by Chomsky ( 1957) . The transformation
rules differ from the phrase-structure rules in that an entire string
of symbols (sentence) may be rewritten by application of a particular
transformation. However, not all transformations can be applied to

all strings.

The application Of the passive transform, for example,
to a symbolic statement, NP1 -- V --- NPZ, allows for rewriting as

NPZ -- be —- V + en -- by + NPl. That is, "The boy ate the apple, “
may be rewritten as, "The apple was eaten by the boy. “

In the grammar prOposed by Chomsky, phrase-structure
rules apply to the initial symbolic constituents (e.g. , NPl) and to
the products of the rewriting Of constituents. Transformation rules,
then, apply to the products of phrase-structure rules and sometimes
to the products Of other transformation rules (e. g. , the passive
transform may be applied to a negative sentence) .

Chomsky has suggested that all syntactic transforms
have identical derivation (the kernel string) and thus all transforms
Of a sentence are related. "The boy caught the fish“ is exemplary of
a kernel. Three transformations may be applied to such a kernel,
producing the passive (P), the negative (N), and the question (Q) .
The transformation rules may also be applied to the products Of
transforms, resulting in the passive -negative (PN) , the passive-
question (PQ) , the negative -question (NQ) , and the passive -negative-
question (PNQ) . Miller ( 1962) presented a cubal representation of
the transformational relations (Figure l) . Each vertex represents
a type of sentence structure; adjacent vertices differ in only one
transform. Several psychological investigations are premised on

the notion that transformationally related sentences are psychologically

related. Sentences differing in only one transform should be more

closely related than sentences differing in two or more transforms.

 

Figure 1. Syntactic relations Of the eight sentences.

Behavioral measures Of the relationship among these
transforms have been reported by Miller ( 1962) , Miller and McKean
(1964) , Clifton, Kurcz, and Jenkins (1965) ,‘ and Clifton and Odom (I966) .
The evidence indicates that there is some correspondence between
transformation relations and the behavioral measures. For example,
using sentence identification time as a measure Of relationship,

Miller reported that the time taken to match up two sentences which
differ only in transform was a function of the number Of transfor-

mation rule 5 applied .

Miller and McKean recorded the time required for actually
transforming sentences. The results indicated that active-affirma-
tive (K) sentences may be psychologically primary and easiest to
process. Formation of the passive took more additional time than
formation Of the negative. When both transformations were required,
the additional time needed for transformation was approximately the
sum of the times for making two separate transforms.

Another behavioral measure of relationship has been em-
ployed by Clifton, Kurcz, and Jenkins. The degree to which a motor
reSponse generalizes between sentences as a function of transforma-
tional relatedness was explored. _S_s saw a list of sentences for four
trials. They were instructed to press a key immediately after read-
ing each sentence. Then a longer list was presented and §S were
instructed to press the key only when they recognized a sentence
from the first list. Sentences closely related transformationally
were confused more than sentences having no grammatical relation,
supporting the notion that sentences that are analyzed as being closely
related grammatically are psychologically similar. Additional sup-
port was reported. Sentences differing by a combination of two
transforms produced less generalization decrement than sentences
differing by only one of the two transformations. The authors caution
that generalization of semantic and physical similarity rather than

syntactical similarity may account for the results.

Miller has Offered an hypothesis concerning the manner
in which transformed sentences are acquired. This hypothesis
assumes that every sentence of a language could not possibly be
memorized since memory capacity is limited, and that the psycho-
logical processing Of sentences corresponds to the rules Of genera-
tive grammar. It is hypothesized, in accord with the transformation
rules, that sentences are remembered as kernel plus a "footnote"
about the transformation.

Evidence for such a two stage process in the memori-
zation of sentences has been reported by Mehler and Miller (1963) .
Using a retroactive interference paradigm, it was demonstrated
that sentence content was first learned (and was first subject to
interference), then syntactic details. Semantic aspects appeared
to be coded separately from syntactic details. Similarly, Mehler
( 1963) , using a method Of prompted recall for sentences written
in the active, passive, affirmative, negative, declarative, or
interogative, reported that most errors were due to syntactical
confusions. An explanation similar to Miller's was Offered:

". . . §S analyze the sentences into a semantic component plus

syntactic corrections when they learn them, and . . . this separa-
tion Of semantic content from syntactic form is one reason that the
general meaning of a message is generally so much easier than its

exact wording . "

10

As a test of Miller's hypothesis, Savin and Perchinock
( 1965) measured the amount of memory capacity remaining after
memorization Of a sentence in each of 11 structural forms. The
results supported the hypothesis: (a) the kernel sentence was
found to require the least memory space; (b) transformed sen-
tences somewhat more; and (c) multiple transformations always
required more Space than singular transformations. The under-
lying notion was that more memory Space should be required if
the transform phrase markers were encoded separately from the
kernel string.

Gough ( 1965) used a procedure in which _S_s were re-
quired to decide whether a sentence was true or false with respect
to a picture. Active sentences were verified more quickly than
passive, affirmative more quickly than negative ones. The re-
sults were interpreted as indicating that §_s must transform a sen-
tence into its kernel before it is understood.

The Mehler and Miller, 'Mehler,_.5avin.and Perchinock, and
Gough studies lend indirect support to Miller's hypothesis about the
way in which transformed sentences are learned or processed, i.e. ,
a two stage process, kernel plus "footnote. "

The results of one study investigating the nature Of the
words used by _S_s as actor, verb, and Object in passive and active

sentence frames was interpreted as negative evidence for the two

11

stage hypothesis (Clark, 1965) . The pattern of uncertainty in the
actor, verb, and Object, as well as the use Of animate or inanimate
nouns as actors and Objects differed in the two syntactic forms.
The author suggests that the results point to a sequential left-to-
right generation of passive sentences rather than the application Of
the passive transform to active sentences. Because Of the incom-
parability of what was required Of the _S_s in this experiment with

the memory experiments, it is difficult to evaluate Clark's evidence

with respect tO the hypothesis.

The Present Experiment

 

The differential inﬂuence of syntactic structures (active,
passive, negative, question, etc.) on psychological functioning has
been reflected in behavioral measures such as reaction time, motor
response generalization, recall, affirmation time. The experimental
tasks employed to Obtain these measures have required that S
memorize or give a simple reaction, with the exception of the Miller
and McKean study in which _S_s were asked to make a particular trans-
formation. It was the purpose Of the present experiment to further
investigate the inﬂuence of syntax on psychological functioning
(Operations Of a language user) through the use Of a syllogistic
reasoning task. This task required §S to formulate a sentence in

solution to some Of the problems rather than to memorize sentences.

12

The task allowed for four behavioral measures of the inﬂuence of
transforms on syllogistic problem solving: latency, perceived
difficulty, number Of currect solutions, and the transforms of the
verbal responses given as solutions. Of particular interest was
the effect Of the application of the Passive transform to simple
Active -Affirmative statements comprising the major premises Of
the syllogisms. Interest was centered on the Passive transform
since there are fewer contextual variables such as truth and false-
ness or exceptionality (Wason, 1965) , which might interact with
the structural manipulation. The Negative transform was of
secondary interest only, for that reason. The nature of the syllo-
gism did not permit the construction of the question and emphatic
transforms.

It is to be emphasized that it was not the purpose Of this
study to investigate the inﬂuence Of syntax on how peOple think, but
to investigate the influence Of syntax on the operations of the language
user. Miller ( 1949) has stated emphatically that the laws Of logic
are Simply not the laws of thought. The syllogism has proven a
useful tool for the demonstration of the Operation of verbal habits
in problem solving and was employed as such in the present experi-
ment. Deviations from the typical formal syllogistic reasoning task
were necessary: (a) problems were written in content rather than

symbolic form; (b) a transitive verb was employed in the major

l3

premise to allow for the Passive transform; and (c) _S_s had to for-
mulate a response rather than choose from alternatives.

To aid in the understanding of the experimental task
the following description of syllogisms, adopted from Chapman and
Chapman (1959) , is presented.

A syllogism consists of three statements, the first two
Of which are premises, the third a conclusion:

1. The major premise states the relationship between the
middle term (common to both premises) and the predi-

cate of the conclusion.

2. The minor premise states the relation between the middle
term and the subject of the conclusion.

3. The conclusion is an inferred, or deduced relation be-
tween the minor term (subject) and the major term
(predicate) .

There are four categorical propositions used in expressing the three

statements of a syllogism:

 

Name Expression Smbol
Universal affirmative All SS are PS A
Universal negative NO 88 are Ps E
Particular affirmative Some 83 are Ps 1
Particular negative Some 83 are not PS O

Each Of the three statements Of a syllogism could be written as
A, E, I, O. The combination in a given problem determines its
mood. There are 16 possible combinations of major and minor

premises, most Of which yield nO valid conclusion. An example Of

14

a valid syllogism of the All mood in symbolic logic and content form

is as follows:

All Ms are PS If all carnivores eat meat,
Some SS are Ms And some animals are carnivores;
Therefore: Then,

Some Ss are Ps Some animals eat meat.

An example Of an invalid syllogism of the III mood is as follows:

Some Ms are PS If some typists transcribe shorthand,
Some SS are Ms And some secretaries are typists;
Therefore: Then,
Some 55 are PS Some secretaries transcribe
shorthand.

The use of the syllogistic reasoning task was chosen
since it allowed (a) for disguising the nature Of the inquiry from S5;
(b) for manipulation of the syntax of the material to be read by S8;
(c) for a verbal reSponse from SS, but still constraining responses
enough to be analyzable; and (d) because Of the constraint, allowed
for analysis of tendencies to reSpond to the atmosPhere created by
the syntactic mood of the problem.

If the Active ~Affirmative is the psychologically primary
structural form Of semantic content, it may be hypothesized that
when information which must be related to other information is
stated in the Passive (or Negative), it will require more time to
relate the material (reach a solution) . This alteration would also

be likely to be considered more difficult by SS. If it is true that

15

SS transform a sentence into their own words or the kernel form,

it would be predicted that _S_s will phrase their responses predomi-
nantly in the Active irrespective Of the transform of the premises
and that perhaps the extra time SS take in responding to transformed
sentences is because of the Operation Of “putting into own words. "
(Mehler, 1963; Gough, 1965.)

Another consideration may determine the structrual
form of the response given by an S. It was mentioned earlier that
Ss' solutions to syllogisms tend to be revealing Of verbal habits
used as a substitute for thinking (through error analysis) .

Many of the errors that SS make in formal syllogistic
reasoning have been interpreted as the result of atmosphere effect
(Woodworth ¢ Sells, 1935; Sells, 1936; Chapman ﬁ Chapman, 1959;
Simpson f Johnson, 1966) . The Operation of atmosphere effect in
formal syllogistic reasoning has been recently stated as an hypothe-
sis and support for it found: "When both premises of a syllogism
contain the same qualifier, 'are' or 'are not' or the same quantifier,
'all' or 'some, ' many SS will accept a conclusion which also contains
the common term. " (Simpson ¢ Johnson) Atmosphere effect has
also been described more generally as a temporary set of an indi-
vidual arising from a given situation or problem in which there are
a limited number Of response alternatives. This set leads the
individual to respond in keeping with the general mood or tone of the

situation or problem (Sells) .

16

It was suggested that SS would probably formulate a
solution in the Active form in the present experiment, irrespective
of the transform of the premises in keeping with the findings Of the
current literature. However, it might be possible to expand upon
the definition of atmOSphere effect and propose that SS will formu-
late a solution in keeping with the structural form--active or
passive--Of the major premise as well as in keeping with the quali-
fiers and quantifiers.

Both predictions may be plausible. The structural form
that an S actually employs may depend in part on the difficulty of a
given Passive problem for him. SS may find it "necessary" to adOpt
a set consistent with the syntactic mood on the difficult problems in
formulating a response, i.e. , change strategy and respond in the
Passive after rote substitution of the terms. With less difficult
problems SS may find it possible to actively attempt to relate the
information given in the major premise to that given in the minor
premise in reaching a solution. For the easier problems it might
be predicted that solution would be most Often in the Active form in
accord with the first hypothesis.

Thus, the major questions Of concern were:

1. Does it take SS longer to solve problems when they are
written in the Passive (or Negative) transform than

when they are written in the Active (or Affirmative) .

2. DO SS consider problems more difficult when they are

written in the Passive (or Negative) transform than
when they are written in the Active (or Affirmative) .

17

What is the predominant verbal response form--active
or passive. DO Ss have a tendency under any conditions
to respond in keeping with the transform Of the premises.

What might account for the additional solution time for
problems written in the Passive transform, if that
occurs.

METHOD

Subjects

The SS were 64 students enrolled in an introductory psy-
chology course at Michigan State University. Course credit was

given for participation .

Materials

Eight combinations Of universal and particular, affirma-
tive and negative premises were used in preparing the syllogisms:
AA, AI, EA, El, II, IA, AE, IE. In symbolic form, A = All X is Y;
E = All X is not Y; I = Some X is Y. Four instances Of each of the
eight combinations Of premises were included, making a total of 32
syllogisms. A valid conclusion could be reached for each Of the
first four combinations of premises listed above (determinant syllo-
gisms); no valid conclusion was possible for any of the other four
combinations (indeterminant syllogisms) .

The four determinant combinations of premises were
balanced in terms of the use of universal (all) and particular (some)
quantifiers, and negative terms (not). The four indeterminant com—

binations were balanced similarly except for the substitution Of II

18

19

(two particular premises) for AA (two universal premises). The
indeterminant combinations were included primarily as buffer items
to reduce _S_S' response set tendencies. The major premise of each
syllogism was written with a transitive verb, the minor premise
always with ”are. " Half (8) Of the determinant syllogisms contained
a negative term in the major premise and no negative term in the
other half. Half of the indeterminant syllogisms contained a nega-
tive term in the minor premise and no negative term in other other
half. The semantic content was different for each of the 32 syllo-
gisms.

Voice (Active, Passive) and Position (1, 2) of the major
premise Of each syllogism was varied. Rather than designating a
particular syllogism to represent a particular combination of
Voice and Position, four forms (A, B, C, D) were constructed in
order to counterbalance semantic content. The major premise
of each syllogism was written four ways: Active, first Position;
Active, second Position; Passive, first Position; Passive, second
Position. Each of the four variations of a syllogism was systemati-
cally assigned to a different form. Thus, each of the 32 syllogisms
appeared on each of the four forms but represented a different com-
bination of variables (Voice and Position) . Each form contained

four examples Of each Of the four combinations of variables for

20

determinant items and four examples Of each combination for in-
determinant items.

The two premises for each syllogism were typed on a
4 x 5 plain white card. An 11-point difficulty rating scale was
printed on a 5 x 8 white card. The left Side was labeled EASY,

the right Side HARD.

Procedure

 

_S_ was seated at a table at right angles to E_. The in-
structions were read and explanatory diagrams drawn for three
practice problems in symbolic logic form. SS were informed that
their solution time would be recorded but that the task was not a
speed test. The instructions cautioned against reliance on know—
ledge Of the semantic relationship among the terms of a problem.

SS were asked to pay attention to the logical relationship established
by "all, " "some, " and "no, " as illustrated in the practice problems
where A, B, and C, having no meaning were used. They were then
informed that E . would give them the two premises for each prob-
lem on a card. After reading the premises and reaching a conclu-
sion, they were to tell it to E. If they thought that there was no
conclusion that followed, they were instructed to say "no conclusion. "

The stimulus cards were placed face down on the table
in front Of E. The cards were handed to S one by one. E saw only

the code number of the card handed to S to minimize possible E bias.

21

E started the stOpwatch as soon as Stook the card. SS promptly
concentrated on each card. The stOpwatch was stopped at the first
word of S's response. The verbal response and solution latency
were recorded for each syllogism. _S_ rated each problem on the
11-point scale for difficulty after each solution.
Four sequences of syllogistic content were used for

each form with four SS assigned to each sequence. The sequences
were:(a) problems 1-32; (b) 32-1; (c) 16-1, 32 -17; and (d) 17-32,

1-16.

Expe r imental De 3 ign

 

The experimental design for determinant items was a 2 x
2 x 2 factorial with repeated measures. The design for indeterminant
items was also a 2 x 2 x 2 factorial with repeated measures. The
independent variables were: Voice (Active, Passive); Qualification
(Affirmative, Negative) ; and Position Of the major premise ( l, 2) .

Four dependent measures were Obtained: solution latency;
difficulty ratings; transforms of SS verbal solutions; and number Of

correct solutions .

RESULTS

Solution latency and difficulty rating data were treated
similarly and will be considered together wherever possible to
di5p1ay consistencies in the trends. The 153.1 coefficient of agree-
ment in rank order was computed for the eight latency and rating
means (representing each combination of variables) . Latency and
rating scores were found to be in high agreement (Egg = . 93) .

Table 1 presents the mean latencies and ratings for the
various transforms in order of increasing magnitude (collapsed
across Position). SS required less time to reach a solution when
the major premise was written in the Active rather than Passive,
and when written in the Affirmative rather than Negative. Corres-
pondingly, the Active problems were rated easier than Passive ones,
and Affirmative problems easier than Negative ones.

Separate statistical analyses were performed on deter-
minant and indeterminant latencies and ratings for three reasons:

( l) confounding due to the difference in location Of the Negative
term (major premise for determinant items; minor premise for
indeterminant items); (2) the imbalance in the use Of particular

(some) premises in constructing the materials (this imbalance

22

Z .3

would be expected to reduce differences in an Affirmative-Negative
comparison across all items, and to enhance the difficulty of the
indeterminant items); (3) the difference in the type Of correct
solution ("no conclusion" versus a valid statement of relationship).

It would have been possible to balance the use Of parti-
culars, universals and negatives—-but only at the expense Of in-
creasing the liklihood Of a response set tendency which would have
been incompatible with the purposes Of theexperiment. The deter-
minant-indeterminant comparison was not critical to the major pur-
poses. Consistencies in the results with determinant and interminant

items may be considered as indicative of the generality Of the results.

Solution Latency

 

Significant Voice, Affirmative-Negative, and Position
main effects were found for both determinant and indeterminant item
solution latencies. Latency was augmented when (a) the major
premise was written in the Passive transform, (b) the major or
minor premise was Negative, or (c) the major premise was in the
second Position on the stimulus card.

The Voice x Position and the Voice x Position x Affirmative-
Negative interactions reached significance for determinant item laten-
cies. Inspection of the means indicated that the significant Voice x

Position interaction may be the result of disproportionately long

latencies when a Passive premise appeared in the second Position.
The three-way interaction is presented graphically in Figure 2.
A Newman-Keuls multiple comparison revealed that placing the
major premise in Position 2 significantly (p < .01) increased Passive-
Negative latencies only. Position did not significantly inﬂuence other
variations. (p > .05)

The multiple comparison also revealed that Active-
Negative and Passive-Affirmative mean latencies were not signifi-
cantly different (p > .05) in either Position. Mean latencies for
all variations were significantly different from Active -Affirmative
latencies with the exception Of the Passive -Affirmative in Position 1,
and Active-Negative in Position 1 (from Active -Affirmative in Posi-

tion 1 only).

Difficulty Ratings

 

The results Of the analyses of variance of difficulty ratings
are summarized in Table 4 and Table 5. Voice, Affirmative-Negative,
and Position main effects for determinant item ratings were signifi-
cant and in the same direction as for latencies. Items were rated
more difficult when (a) the major premise was written in the Passive
transform, (b) the major premise was Negative, or (c) the major

premise was in the second Position.

25

Voice and Affirmative-Negative main effects for indeter—
minant items reached significance. The Position main effect failed
to reach a statistically significant level. Indeterminant items were
rated more difficult when (a) the major premise was written in the
Passive, or (b) the minor premise was Negative.

The significant Voice x Position x Affirmative-Negative
interaction for determinant item ratings is illustrated in Figure 3.

A Newman-Keuls multiple comparison revealed that Position 2 had
a significant (p > .01) influence on Passive-Negative ratings only.
Position did not significantly inﬂuence other variations. (p > .05)

The multiple comparison also revealed that Active-
Negative and Passive --Affirmative items were not rated significantly
different (p > .05) in either Position. Mean ratings for all variations
were significantly different from Active-Affirmative ratings with the
exception of the Passive -Affirmative in Position 1 and 2 (from Active-

Affirmative in Position 2 only) .

Summary of Analyses of Variance

 

A summary of the results of the four analyses Of variance
is presented in Table 6. The significant main effects were all in the
same direction: Active < Passive; Affirmative < Negative; and

Position 1 < Position 2.

l\)
0\

Correct Solutions

 

The number of correct solutions was tabulated for Active,
Passive, Affirmative, and Negative items. Determinant items were
scored as correct only if a valid conclusion was given. Indeterminant
items were scored as correct only if "no conclusion" was given as a
response. Table 7 presents the mean number correct (out of pos-
sible 4) . The reversal in order of magnitude Of the means for de-
terminant and indeterminant items should be noted. This reversal
may be due to the difference in the form Of the response necessary
to be considered correct: a valid conclusion in sentence form as

Opposed to "no conclusion. "

Transform of Verbal Solution

 

The number of responses given in the Active, Passive,
Affirmative, and Negative transforms or as "no conclusion'l were
tabulated for each combination of variables represented within
premises. Percentages were found and are presented in Table 8.

The percentages in identical premise transform-response transform
cells are underscored. For both determinant and indeterminant

items, a considerably greater percentage Of responses was given

in the Active (85. 5% and 61. 2%) than in the Passive (11.1% and 15. 9%).
Position had little inﬂuence on the structure of the response given.

The increased percentage of "no conclusion" responses to Negative

premises for both determinant and indeterminant items should be noted.

27

Even though responses were given predominantly in the Active,
a two-category chi-square analysis revealed that sifnificantly more than
half of the SS did give atleast one Passive response (I2 = 5. 06, dfl, p < . 05) .

A median test was performed to determine if there was any
relationship between the number Of Passive reSponses Ss gave to Passive
items and Speed of solution. SS were categorized as falling above or be-
low the median solution latency score (70 seconds) for Passive determin-
ant items and above or below the median number of Passive reSponses
given (1) . A nonsignificant x2 = 0. 399, (if 3, was Obtained.

To determine if a differential amount of time might be required
to reSpond in the Active or Passive, a finer measure was employed. The
mean solution time was found for determinant Passive problems to which a
Passive reSponse was given; likewise, for those to which an Active reSponse
was given. A t_-test for related measures was performed and a nonsignifi-
cant L = 0.09 (df 39) was found. A nonsignificant tendency (_t_ = l. 91, df 39,
p < .10) was found for SS who gave at least one Passive response to deter-
minant Passive items, to rate those problems to which they gave a Passive
response as more difficult than those to which they gave an Active response.

Two opposing hypotheses were made concerning the form of
the reSponse Ss might give and the inﬂuence Of latency. The nonsignificant
results above may in indicative Of the Operation of atmOSphere, on the one
hand, increasing latency of reSponses given in the Passive, and on the
other, of SS avoiding the time consuming Operation of putting the premise

in their own words and reSponding directly in the Passive.

28

When SS were divided into two groups according to whether
they had a plus score (took more time to reSpond with a Passive than
Active verbal reSponse) on the amount Of time required to reSpond to
Passive determinant items with a Passive, and those who had a minus
score (took less time to respond with a Passive than Active), it was
found that SS with plus scores gave significantly more Passive reSponses
than did those SS with minus scores (_E_ = 2. 90, df 39, P < .01) . The S3
with plus scores also rated Passive problems to which they gave a Pas-
sive response an average of l. 08 m difficult than those to which they
gave an Active response, while SS with minus scores rated Passive
problems to which they gave a Passive reSponse an average of . 272
£83. difficult than those to which they gave an Active reSponse.

This dichotomy may represent those SS who did not give
the Passive as an atmOSphere response, SS who did not give the Pas-
sive as an atmOSphere reSponse taking less time to give Passives
than Actives to Passive problems.

To determine if SS, in general, responded to the atmos-
phere of problems, a L—test for related measures was performed.

SS gave significantly more Passive reSponses to indeterminant items
than to determinant items (_t_ = 3.46, df 63, p < .01) . It might be
noted that those SS who were in the plus group above gave an average

of 1. 14 more Passives to indeterminant items than did minus SS.

29

Table 1. Mean solution latency and difficulty ratings for active -affirma-
tive, active -negative, passive-affirmative, and passive «negative items.

 

 

 

 

 

Latency Rating
Transform
Determinant Indeterminant Determinant Indeterminant

Active -

Affirmative 6. 30 9. 89 2. 41 3. 50
Pas sive-

Affirmative 8.10 11.86 2.82 3.99
Active-

Negative 8.30 14.61 . 3.03 4.57
Passive -

Negative 11.03 15.69 3.63 4.81

 

Table 2. Summary of the analysis of variance of determinant-item
solution latencies.

 

 

 

Source df Mean Square F P
Active-Passive (V) 1 2, 619. 07 39. 28 . 01
Affirmative-Negative (N) 1 3, 120. 50 46. 80 . 01
Position (P) 1 661.13 9.91 .01
Vx P 1 343.86 5.14 .05
Nx P 1 149.50 2.24 —-
VxN 1 108.78 1.63 --
VxPxN 1 712.56 10.69 .01
Subjects 63 385. 30

Error 441 66.68

 

Total 511

 

30

Table 3. Summary of the analysis of variance of indeterminant-item
solution latencies.

 

 

 

 

Source df Mean Square F P
Active-Passive (V) l 1, 194. 38 8. 27 . 01
Affirmative-Negative (N) 1 9, 358. 83 64. 77 . 01
Position (P) 1 686.81 4.75 .05
VxP 1 157.83 1.09 .--
N x P j 1 13.46 -- --
V x N l 101. 53 -- --
VxPxN 1 272.63 -— --
Subjects 63 1, 745.03

Error 441 144. 50

Total 511

 

Table 4. Summary of the analysis of variance Of determinant-item dif-
ficulty ratings.

 

 

 

Source df Mean Square F P
Active-Passive (V) 1 131.02 41.46 .01
Affirmative-Negative (N) 1 257. 36 81.44 . 01
Position (P) l 19.93 6.31 .05
V x P 1 1.06 —- --
N x P 1 5.08 1.61 --
V x N l 4.69 1.84 --
VxPxN 1 30.72 9.72 .01
Subjects 63 33.87

Error 441 3.16

Total

 

511

 

31

Table 5. Summary of the analysis of variance Of indeterminant-item
difficulty ratings.

 

 

 

 

Source df Mean Square F P

Active-Passive (V) 1 68.84 9.30 .01
Affirmative-Negative(N) 1 451.87 61.06 .01
Position (P) 1 21.53 2.91 --
V x P 1 8. 99 1. 21 --
N x P l 7.27 -- --
V x N 1 6.26 -- --
VxPxN 1 25.55 3.45 --
Subjects 63 49. 99

Error 411 7.40

Total 511

 

Table 6. Summary of results of latency and rating analyses of variance.

A L

 

Latency Rating

 

 

Source
Determinant Indeterminant Determinant Indeterminant

 

V01 C e g. >.'= ::: ::< :2: >k * *

Affirmative -
Negative

PO S it ion >1< >l< ..< =:: _
V x P * — - ..
N x P - j - .. -
V x N - — - ..
V X P X N *‘i‘ - :1: _

 

>:< P < .05
p < .01

32

Table 7. Mean number Of correct solutions.

 

 

 

, Determinant Indeterminant
Premise
Items Items
Active-Affirmative 3. 23 O. 39
Passive-Affirmative 2. 89 0. 48
Active -Negative 1. 8 9 1. 56

Passive-Negative 1. 58 1. 25

 

33

Table 8. Percentage of responses given in Active, Passive, Affir-
mative, Negative, NO Conclusion for each combination Of variables.

 

 

 

 

 

 

 

 

 

Responses
ACTIVE PASSIVE
NO
. . . . . , Conclusion
affirmative negative affirmative negative
Determinant-
Item Premise
ACTIVE -aff 1 99.2% 0 % 0 070 0 "/0 0.8 070
aff2 96.9 0.8 0.8 0 1.5
neg 1 3.1 92. 2 0 0 4. 7
neg 2 5. 5 87. 5 0 0 7. 0
PASSIVE- aff 1 82.8 0 17.2 0 0
aff2 81.2 0 17.2 0 1.6
neg 1 3.1 68.8 0.8 25.0 2.3
neg 2 3.1 60.2 2.3 25.0 9.4
%Of Total 46.8 38.7 4. 9 6.2 3.4
Indete rminant -
Item Premise
ACTIVE -affl 90.6 0.8 0 0 8.6
aff 2 86.0 3.1 0 O 10.9
neg 1 10.2 51.5 0 0 38.3
neg 2 4.7 53.1 1.6 0.8 39.8
PASSIVE -aff1 55.4 0.8 30. 5 0.8 12. 5
aff 2 53.9 0.8 33.6 0.8 10.9
neg 1 4.7 36.0 3.9 23.4 32.0
neg 2 4.7 32.8 3.9 28.1 30.5
% Of Total 38.8 22. 4 9. 2 6. 7 22. 9

 

 

Mean Solution Time (seconds)

 

34

 

 

 

 

o—o active-affirmative
0—0 active-negative
30 - D———CI passive-affirmative
I—-—. passive-negative
25 ., / /'
/
/
/
/
/
/
/
/
/
20 .. / /
./
..——-- ‘ﬂ- 4
nun—- _— ‘-__ "— ‘-.
F— , _— _—__ ___,—-—- ""—
15 ‘ D
__o
C—
10 -
1 _ . 2
POSition
Figure 2. Mean determinant-item latency as a function of Voice,

Negation, and Position.

Mean Difficulty Rating

35

active -affirmative

 

0—0
H active-negative
D---D passive-affirmative
l—--I passive-negative
8.0 ‘
/A
/
7.5 « / /
/
/
/
/
7.0 1 ///
/
V
6.5
6.0 P
D—-——-———._._____ D
5.5
5.0 i
4.5 ‘
4.0 ‘

 

 

Position

Figure 3. Mean determinant-item difficulty ratings as a function of
Voice, Negation, and Position.

DISSC USSION

In general the results suggest that the Active -Affirmative
is the habitual structural form of semantic information for the language
user. SS required less time to solve problems when the information
was in the Active ~Affirmative structural form, found them less dif-
ficult, and made fewer errors. Solutions were most Often given in

the Active irrespective of the transform of the premises.

Solution Latency

 

The determinant-item results indicate that (a) when the
major premise of a syllogism is expressed in the Passive transform,
more solution time is required; (b) when the major premise Of a
syllogism is eXpressed in the Negative transform, more solution
time is required; (3) Passive-Affirmative and Active -Negative prem-
ices lengthen solution time almost equally. Furthermore, the sig-
nificant interaction between Voice, Position, and Affirmative-
Negative indicates that disproportionately more solution time is
required when the major premise is written in the Passive-Negative
and is in Position 2 on the stimulus card.

These results are generally in keeping with those re-
ported by Miller and McKean (1964) and by Gough ( 1965) , also using

36

37

latency measure but quite different tasks. Miller and McKean re-
quired SS to transform one type of sentence into another and
measured the amount of S controlled presentation time. More
time was required when transforming a passive sentence than when
transforming an active (about 1 second); about 0.4 seconds more
was required when transforming a negative sentence than an af-
firmative. The difference between active-passive times was
greater than between affirmative ~negative. When the sentence to
be transformed was both passive and negative, the additional time
required was approximately the sum of the time needed for the two
separate transforms. Time differences were a function Of the
structural form of the sentence being transformed, not the trans-
formation being applied. Thus transformation time was found to
vary with the kind and number of transformations.

Gough reported results indicating that speed Of under—
standing sentences also varies with the number and kind Of trans-
formations that have been applied to the kernel. Active sentences
were understood faster than passive, affirmative faster than neg-
ative ones. Passive sentences were understood faster than negative
ones. The results were interpreted as evidence for the notion that
Ss must transform complex sentences into the kernel before under-
standing. Gough pointed out that transformational complexity was
confounded with frequency and length. This confounding is common

to almost all experiments on syntax.

38

These two experiments were concerned in part with the
differential inﬂuence Of the number and kind of transformations on
latency as well as the influence on the passive compared with the
active, negative compared with the affirmative. This concern is
based on the notion that transformationally related sentences are
perhaps psychologically related. It might be noted that in the Miller
and McKean experiment the negative required less time than the
passive; in the Gough experiment the negative required more time
than the passive. This seems to indicate that the psychological
relationship Of transforms varies with tasks, the sentential mate-
rial (context), etc.

Little can be concluded with respect to the relation Of
the Negative, Passive, and Passive-Negative transforms from the
present experiment either. The Passive AAffirmative and. Active-
Negative premises lengthened solution time almost equally in de-
terminant problems, which might be considered as evidence for
the notion that all transforms equally influence behavioral measures.
It is difficult to determine whether the additional time required for
solution of a problem written in the Passive-Negative was the sum
or greater than the sum of the additional time required for the Pas-
sive and Negative alone because of the significant Position effect.
However, a roughly additive relation was found.

The reason for the diSprOportionately long latencies

when a Passive -Negative problem was in the second Position is

39

not immediately evident especially Since Position had no effect on
other variations. Position of the major premise was varied in the
experiment Simply to insure that if SS read the premises in se-
quence the last premise read would not always be a simple "are"
sentence which might influence the structural form of the response
given. Position was found to have little influence on the form given.
The nature of the inﬂuence Of Position on Passive—Negative latency
may be inferred by comparing the effect Of Position on Ss' task. If
it can be assumed that SS do put the Passive-Negative structured
information into their own words in Active form, and that this Op-
eration requires a little time, it may be hypothesized that when the
Passive-Negative premise is first, the "putting into own words"
Operation may occur simultaneous with the reading Of the second
premise. When the Passive—Negative premise is second, however,
extra time may be needed to perform this Operation before an at-
tempt to relate the two premises is made. The presence Of two sets
Of phrase markers (be-en-by; not) serve to make the sentence
structurally complex and to delay the other operations.

A companion explanation may be that it is more difficult
to identify the middle term, subject of the conclusion, and predicate
when the premise carrying the predicate is the second piece Of in-
formation given and when changes have been made in phrase structure

displacing the habitual subject-Object positions and add another term

40

(not) to the predicate. The perceptual difficulty in identifying the
terms may be augmented by the nature Of the information given in
the two premises. The major premise sets the bounds of the final
reSponse (Cochroaches are not caged by all zoos). The minor
premise is more global (Some menageries are zoos). Knowing

the bounds first may aid in solution when the terms are not immme-
diately discriminable.

The results with indeterminant items indicate that (a)
when the major premise of a syllogism is expressed in the Passive
transform, solution time increases; (b) when the minor premise Of
a syllogism is expressed in the Negative transform solution time
increases; (c) the Negative premise increases solution time (almost
2 1/2 times) more than the Passive.

It would be expected that the indeterminant items would
require more evaluation time in general than determinant items
since the tentative responses might not "sound just right". How-
ever, it is noteworthy that the inﬂuence of transform (Passive and
Negative) on solution time was increasing as with determinant
items. The somewhat greater magnitude of additional time re-
quired for the solution when the minor premise was expressed in
the Negative may reﬂect the Operation Of having to add the Negative
term to the predicate of the major premise in forming a tentative

solution. The Negative always occurred in the major premise Of

41

determinant items SO this Operation was not necessary, as re-
flected in the relatively Shorter additional time needed. It is also
possible that the extra time required for problems when the minor
premise was Negative was a function Of context. There is some
evidence that context influences Speed Of response to negatives
(Gough, 1965; Wason, 1965) . The major premise was a simple
statement Of fact--All vegetarians do not eat meat--whereas the
minor premise was a statement of relationship of a smaller cate-
gory to a larger category-- Some animals are not vegetarians. That
this latter context may be much more difficult is supported by the
difficulty ratings and by comparing the relative differences in la-
tency and ratings between Passive and Negative determinant and
indeterminant items. Passive and Negative determinant problems
took almost equally long to solve while Negative indeterminant pro-
blems took almost three seconds longer to solve than did Passive

problems.

Difficulty Rating_S_

 

The difficulty ratings of determinant items indicate that
(a) when the major premise Of a syllogism is written in the Passive
transform the problem is considered more difficult than when in
the Active; (b) when the major premise Of a syllogism is written in

the Negative the problem is considered more difficult than when in

42

the Affirmative. These results support Miller's contention that the
Active-Affirmative sentence is the habitual syntactic form for SS.
Semantic content exPressed in the habitual syntactic form is easier
to understand (Gough) and thus solving a problem should also be
easier. The rating results substantiate findings with various be-
havioral measures which purportedly reflect the influence Of syntax
on psychological Operations. That is, Ss' perception Of judgment
Of the difficulty Of an item was in high agreement with the inﬂuence
of transform on the latency of solution of an item, suggesting the
inﬂuence Of syntax is "real".

The significant Voice x Position x Affirmative -Negative
interaction revealed that the occurrence Of a Passive -Negative pre-
mise in Position 2 was disprOportionately more difficult than other
premises in either Position. This mirrors the inﬂuence found on
latency.

The ratings of indeterminant items indicate that (a)
when the major premise Of a syllogism is expressed in the Passive
the problem is considered more difficult than when in the Active;
(b) when the minor premise is expressed in the Negative the pro-
blem is considered more difficult than when in Affirmative.

These results also support the contention that Active-
Affirmative sentences are the habitual structual form for SS and

thus easiest for them.

43

Correct Solutions

 

The tabulated results indicate that (a) more correct
solutions are given when premises are expressed in the Active than
in the Passive; (b) more correct solutions are given to problems
expressed in the Affirmative than Negative. (For determinant
items)

It was expected that errors would increase along with
latency and rated difficulty. The expectation was supported, but
since most errors were Omissions of the quantifiers (some and
all), error analysis was not very revealing of the nature Of the in-
ﬂuence Of syntax on Ss' perception of content. Errors other than
Omissions Of quantifiers were few but were almost all actor-Object
confusions in response to Passive premises. The Object rather
than the actor would be quantified.

The indeterminant item results indicate that (a) more
correct solutions are given to Negative than Affirmative items; (b)
the number Of correct solutions given to Passive items differs only
Slightly from the number given to Active items.

This reversal in pattern with Indeterminant items is
Simply due to the nature Of the response necessary to be considered
correct--"no conclusion". SS had a tendency to say "no conclusion"
to those problems which they rated as most difficult (Negative and
Passive). Either it was more obvious that there was no conclusion

to Negative problems or SS used the response as an “out".

44

It would have perhaps been better had this Option been
omitted in the instructions. The analysis Of the structural form Of
responses was considerably weakened because Of the high propor-
tion Of “no conclusion" responses given to indeterminant items as

an “out".

Structural Form <_)_f_ the; Response

 

The percentages indicate that (a) the Active sentence is
the predominant structural form used by SS in speech irrespective
'Of the structural form Of the stimulus material before them; (b)
Position of the major premise has little inﬂuence on the structural
form Of the response given tO syllogisms. The findings again sup-
port Miller's notion that the Active form is the primary syntactic
form for the language user. The predominant use Of the Active
structure in responding, in conjunction with the consistently shorter
solution time and Ss' difficulty ratings (easier), is convincing evi-
dence Of Miller's notion.

There is some evidence (Mehler ¢ Miller, 1963;
Mehler, 1963; Gough, 1965) that SS reword a transformed sentence
into their own words‘- -an active -affirmative sentence--in the pro-
cess of memorizing or responding to a sentence. It has been sug-
gested that the extra time taken in understanding and responding to
transformed sentences may be in part a function of the time re-

quired for this Operation.

45

That SS responded predominantly in the Active to Pas-
sive premises as well as to Active premises in the present experi-
ment and required more time suggests that SS did reword the
information at some point in the solution process and that this pro-
cess may account for longer latency in solution.

Analysis was made of the structural form Of the responses
given as solutions to Passive problems (since more than half of the
SS did give at least one Passive response) in order to gain more in—
sight into the relation between syntax of response, latency, and
difficulty Of Passive problems.

The number Of Passive responses given was found to
have no relation to the time required for solution. Amount of time
to solve problems by giving an Active versus Passive reSponse was
not found to vary. These findings suggested that even if it could be
hypothesized that two Operations--putting into own words, and then
back into Passive--were involved in responding with the Passive, it
was not reﬂected in latency.

It was found, however, thatadichotomy of SS existed
which was producing no differences in the results above: SS who
required less time to solve problems when they gave their response
in the Passive than in the Active; and SS who required more time to
solve Passive problems when they gave their solution in the Passive

than in the Active. That is, SS who responded directly to the Passive

46

without "putting it into their own words" first and those who per-
formed two Operations in formulating a Passive response, formed
the dichotomy.

Both groups Of SS provide evidence that latency is in-
ﬂuenced by the Operation of ”putting into own words". The grouping
of SS who required less time to solve Passive problemswhen they
gave their response in the Passive than Active, gave significantly
fewer Passive responses and rated the problems as easier to which
they gave a Passive response than those to which they gave an
Active. This suggests that the solution of some of the problems
was simply so Obvious that no rewording was necessary for these SS.

It was hypothesized that SS might tend to formulate a
solution in keeping with the structural form or atmosphere—~active
or passive—-Of the major premise, particularly on difficult problems.
That this was not the case in general was indicated by the overall
high percentage Of responses given in the Active. On the problems
where atmosphere responding would be most expected--indeterminant
items--however, somewhat more Passive responses were given
(significantly more than to determinant items) . The half of the Ss
who required more time to solve Passive problems with a Passive
solution than an Active tended to give more atmosphere responses.
They gave more Passives to indeterminant items than did the "less
time" SS, and gave significantly more Passive responses to deter-

minant Passive items. They also rated the Passive problems to

47

which they gave a Passive response harder than those to which
they gave an Active response. This is consistent with the pre-
diction.

These results suggest that it was necessary for the
"more time" SS to change their strategy and formulate a solution in
keeping with the syntax Of the premises on difficult problems, per-
haps out of cautiousness or uncertainty. It is suggested that for-
mulating a Passive response to a problem with which SS were having
difficulty was not a one-step process but involved "putting into own
words" in an attempt to understand the relation expressed. Then,
not feeling certain about the solution, SS reformulated the solution
by adding the Passive phrase markers.

Negative items could not be profitably analyzed with
reSpect to these considerations Since the tendency to reSpond with
Negative when there is a Negative premise is very high in syllo-
gistic reasoning. The Negative problems did require more time,
were more difficult, produced more errors on determinant items,
and considerably more "no conclusion" responses to indeterminant
items, than did the Affirmative problems. This suggests that the
Negative term inﬂuenced the Ss' perception, understanding, and/or
required them to put the premise in the Affirmative form, then put

the Negative term back in.

48

\

It is proposed that the influence of transformed sentences
on latencyf difficulty and the necessity to reword, is a function Of a
frequency Of occurrence and use variable. Transformed sentences
have simply not been as frequently encountered by Ss. Because Of
the consistent combination of structural and semantic aspects (posi-
tion Of actors, Objects, etc. and phrase markers) in the most fre-
quently encountered Active -Affirmative sentence form, the sentence
may function as an integrated response unit. On the other hand,
because the structural and semantic aspects of transformed sentences
are not as frequently encountered, the sentence may not function as
an integrated response unit. The subject, verb, and Object compo-
nents and their relationship may not be perceptually distinctive in
the less frequently encountered forms. SS may have to go through
the process of identifying these components, during the process,
rewording them, before their syntactically mediated relationship be-

come 5 clear .

SUMMARY

Recent research in verbal behavior has been based on

the assumption that there may be a correspondence between syntactic
relations as described by modern linguists and the psychological Op-
erations performed by the language user. The differential inﬂuence
of syntactic structures (active, passive, negative, question, etc.)
has been reﬂected in behavioral measures such as reaction time,
motor response generalization, and recall with sentential material.
The active -affirmative structural form has been regarded as psycho-
logically primary as well as syntactically primary. Some evidence
suggests that SS reword syntactically complex sentences into the
active-affirmative at some stage in processing.

The present experiment investigated the influence of
syntax on the Operations of thelanguage user by varying the structure
of the major premise of syllogistic reasoning problems. Four be-
havioral measures were used: solution latency, rated difficulty,
number Of correct solutions, and structure Of verbal solution. Of
particular interest was the Passive-Active comparison; of secondary
interest the Negative —Affirmative comparison.

It was hypothesized that (a) problems written in the

Passive should take longer to solve and be rated more difficult than

49

50

those written in the Active; (b) problems written in the Negative
should take longer to solve and be rated more difficult than those
in the Affirmative; (c) verbal solutions should be given predominantly
in the Active form; (d) verbal solutions to Passive problems may be
given in the Passive to difficult problems as an atmOSphere reSponse.
A set of 32 syllogisms, each having a different semantic
content, was employed. A valid conclusion could be reached for half
of the syllogisms (determinant items); no valid conclusion could be
reached for the other half of the syllogisms (indeterminant items).
Eight Of the determinant syllogisms contained a Negative term in the
major premise; eight of the indeterminant syllogisms contained a
Negative term in the minor premise. Voice (Active, Passive) and
Position ( 1, 2) Of the major premise of each syllogism was varied.
Four forms were prepared such that each of the 32 syllogisms ap-
peared on each of the four forms but represented a different com-
bination Of variables (Voice and Position) on each. In this way
semantic content was counterbalanced. The two premises for each
problem were manually presented on 4 x 5 white index cards. S
was handed each of the 32 cards in succession by E. S was instructed
to read the premises, reach a solution, and tell it to E. Solution
latency and the verbal solution were recorded. S rated each problem
on an 11-point difficulty scale printed on a card in front Of him after
giving the solution.

SS were 64 introductory psychology students.

51

The hypotheses were confirmed. More solution time was
required when problems were written in the Passive or Negative than
in the Active or Affirmative. The Passive and Negative problems
were also rated as more difficult. Verbal solutions were given pre-
dominantly in the Active (85% to determinant items and 61% to inde-
terminant items) rather than Passive.(evidence for rewording Of
Passive) . With respect to predictionDa dichotomy of SS was found.
One grouping of SS took less time to give a Passive than Active
response to Passive problems and rated those problems to which they
gave a Passive response as easier. The second grouping Of SS took
more time to give Passive than Active responses to Passive problems
and rated those problems to which they gave a Passive reSponse as
harder. This latter group gave Significantly more Passive reSponses
than did the former. It was suggested that perhaps the dichotomy of
SS represents those SS who gave Passive responses to easy problems
because the solution was so obvious for them that no rewording into
the Active was necessary, and those SS who gave the Passive reSponse
to difficult problems as an atmosphere response out of uncertainty
or cautiousness. Two steps may have been involved in the latter Ss'
formulation of a solution accounting in part for the additional solution

time required--rewording into Active, then back into Passive.

REFERENCES

Chapman, L. J. and Chapman, J. P. Atmosphere effect re-examined.
J. exp. Psychol., 1959, 5:8: 220..»226.

 

Chomsky, N. Syntactic Structures. S'-Gravenhage: Mouton, 1957.

 

Clark, H. H. Some structural properties of simple active and pas-
sive sentences. J. verb. Learn. and verb. Behav., 1965,
1, 365-3700

 

Clifton, C., Jr., Kurcz, I. and Jenkins, J. J. Grammatical rela-
tions as determinants of sentence similarity. J. verb.
Learn. and verb. Behav., 1965, 4_, 112—117.

 

, and Odom, P. Similarity relations among certain English
sentence constructions. Psychol. Monoggn 1966, _8_0,
Whole NO. 613, 1-35.

 

Coleman, E. B. Responses to a scale of grammaticalness. J. verb.
Learn. and verb. Behav., 1965, 3, 521-527.

 

Fodor. J. A. and Beven, T. G. The psychological reality of linguis-
tic segments. J. verb. Learn. and verb. Behav., 1965,
1, 414-420.

 

Gough, P. B. Grammatical transformations and speed Of under-
standing. J. verb. Learn. and verb. Behav., 1965, 1,
107-111.

 

Johnson, N. F. The psychological realtiy of phrase-structure rules.
J. verb. Learn. and verb. Behav., 1965, _4, 469-475.

 

Marks, L. F. and Miller, G. A. The role Of semantic and syntac-
tic constraints in the memorization of English sentences.
J. verb. Learn. and verb. Behav., 1964, S, 1-5.

 

Mehler, J. Some effects of grammatical transformations Of the
recall Of English sentences. J. verb. Learn. and verb.
Behav., 1963, S, 346-351,

 

52

53

, and Miller, G. A. Retroactive interference in the recall
of simple sentences. Brit. J. Psychol., 1964, _SS,
295‘3010

 

Miller, G. A. Some psychological studies of grammar. Amer.
Psychol., 1962, _11, 748-762.

Language and Communication. New York: McGraw-Hill
Book Company, Inc., 1949, Chap. 11.

, and Isard, 8. Some perceptual consequences of linguistic
rules. J. verb. Learn. and verb. Behav., 1963, ._?._,
217-228.

, and McKean, K. A chronometric study Of some relations
between sentences. Quart. J. exp. Psychol., 1964, _1_(1,
297-307.

 

, and Selfridge, J. A. Verbal context and recall Of meaning-
ful material. Amer. J. Psychol., 1950, Q, 176-185.

 

Salzinger, K., Portnoy, S., and Feldman, R. S. The effect of order
Of approximation to the statistical structure of English on

the emission Of verbal responses. J. exp. Psxchol. ,
1962, E, 52-57.

 

Savin, J. B. and Perchonock, E. Grammatical structure and the
immediate recall Of English sentences. J. verb. Learn.
and verb. Behav., 1965, 1, 348-353.

 

Sells, S. B. The atmosphere effect: an experimental study of
reasoning. Arch. Psychol., 1936, 2_9_, 3-72.

 

Simpson, M. E. and Johnson, D. M. .Atmosphere and conversion
errors in syllogistic reasoning. In Press.

Wason, P. C. The contexts of plausible denial. J. verb. Learn.
and verb. Behav., 1965, 1, 7-11.

 

 

Woodworth, R. S. and Sells, S. B. An atmOSphere effect in formal
syllogistic reasoning. J. exp. Psychol., 1935, _1_S,
451~462.

 

APPENDIX A
Instructions

This is a task. in logical reasoning. The problems you
will have are called syllogisms. A syllogism consists of three
statements; the first two are premises which state relationships,
and the third is a conclusion which ”hOpefully” follows from the
premises.

The problems might be in this form (draw Euler diagrams

and explain each premise as draw in terms of relationships.)

If all As are BS

And some Cs are As
Then, some Cs are BS
If no As are BS

And some Cs are As

Then, some Cs are not BS

If some As are Bs

And some Cs are As

Then, this time there is not a conclusion. We do not have
enough information about the relationship between the
Cs and BS this time.

The problems you will have will be similar to these
except that in place Of the AS, Bs, and Cs will be words so that
you will have sentences which make more or less sense. In a way
these will be easier, in a way more difficult. You remember, the
relationship was established by the alls, somes, and nos, not

54

55

whether, A, B, or C meant anything. SO-—you will have to be care-
ful not to rely on your knowledge Of the semantic relationship be-
tween the words-—-~the meanings-~but to pay attention to the logical
relationship established by the alls, somes, and nos.

I will give you the two premises on a card. Read them.
When you have decided what the conclusion is, tell it to me. If you
think there is no conclusion that follows, say "no conclusion. ”

After each problem I want you to tell me how difficult
you thought it was (Show scale). Pick a number down here (point
toward end marked easy) if you thought it was pretty easy and one
up here (point toward end marked easy) if you thought it was pretty
easy and one up here (point toward end marked easy) if you thought
it was pretty easy and one up here (point toward end marked hard)
if you thought it was hard.

I will be timing you but this is BEE a Speed test. Work
at a rate comfortable for yourself.

OK—-let's try the first one.

APPENDIX B

Difficulty Rating Scale

EASY 1 2 3 4 5 6 7 8 9 10 l 1 HARD

56

APPENDIX C
Syllogisms

Determinant items

If all. plants need water,
And all ﬂowers are plants;
Then,

If all teenagers enjoy movies,
And all 15-year-Olds are teenagers;
Then,

If all spiders Spin webs,
And all tarantulas are spiders;
Then,

If all students must take exams,
And all freshmen are students;
Then,

If all clubs elect Officers,
And some groups are clubs;
Then,

If all carnivores eat meat,
And some animals are carnivores;
Then,

If all sweets cause cavities,
And some desserts are sweets;
Then,

If all surgeons perform Operations,
And some doctors are surgeons;
Then,

57

58

If all glass does not conduct electricity,
And all windows are glass;
Then,

If all dogs do not tell storeis,
And all poodles are dogs;
Then,

If all farmers do not raise elephants,
And all husbandrymen are farmers;
Then,

If all grocery stores do not sell furniture,
And all supermarkets are grocery stores;
Then,

If all renters do not pay school taxes,
And some residents are renters;
Then,

If all zoos do not cage cockroaches,
And some menageries are zoos;
Then,

If all newsPapers do not report secrets,
And some journals are neWSpapers;
Then,

If all schools do not require basketweaving,
And some institutions are schools;
Then,

Indeterminant Items

If some women outweigh the average man,
And some women are Spinsters;
Then,

If some gestures express emotion,
And some movements are gestures;
Then,

If some gamblers lose money,
And some Speculators are gamblers;
Then,

59

If some typists transcribe shorthand,
And some typists are secretaries;
Then,

\I

kIf some hunters shoot pheasants,
And all hunters are sportsmen;
Then,

*If some dressmakers use thimbles,
And all dressmakers are seamstresses;
Then,

If some drivers Obey laws,
And all chauffeurs are drivers;
Then,

If some bears raid dumps,
And all grizzlies are bears;
Then,

If all children recite nursery rhymes,
And all children are not poets;
Then,

If all lawyers document cases,
And all bailiffs are not lawyers;
Then,

If all cobblers mend Shoes,
And all cobblers are not factory workers;
Then,

If all lamps provide illumination,
And all chairs are not lamps;
Then,

If some rules guide behavior,
And all rules are not promises;
Then,

If some sentences represent ideas,
And all fragments are not sentences;
Then,

* Error in classification was found; a valid conclusion is possible.

60

If some researchers conduct polls,

And all researchers are not philOSOphers;
Then,

If some rodents carry disease,

And all invertebrates are not rodents;
Then,